Corollary 2: Subsystem Code Parameters #
For the horizontal subsystem balanced product code (Def_29) over an s-regular graph X
with free ℤ_ℓ-action, the code parameters [[N, K, D_X, D_Z]] satisfy:
N = 3|X₁|K = dim H₁(C(X/ℤ_ℓ, L)) ≥ (2k_L/s - 1)|X₁|/ℓD_Z ≥ |X₁| · min(α_ho/2, α_ho·β_ho/4)D_X ≥ min(α_co·|X₁|, α_co·β_co·|X₁|/2, ℓ·α_co/(4s), ℓ·α_co·β_co/(4s))
Main Results #
numQubits_eq— N = 3|X₁| (from balanced product dimension + regularity)subsystemCodeParameters_K— K = dim H₁ʰ = dim H₁(quotient Tanner code) (Thm_12)subsystemCodeParameters_K_lowerBound— K ≥ (2k_L/s - 1)|X₁|/ℓ (Thm_12 + Thm_7)subsystemCodeParameters_DZ— D_Z ≥ |X₁| · min(α_ho/2, α_ho·β_ho/4) (Thm_13)subsystemCodeParameters_DX_combined— combined D_X bound (Thm_14)subsystemCodeParameters_DX_for_nontrivial— combined D_X for any nontrivial class
Abbreviations #
The number of edges (1-cells) of the graph X.
Equations
Instances For
The number of vertices (0-cells) of the graph X.
Equations
Instances For
Step 1: Number of Physical Qubits N = 3|X₁| #
The physical qubits correspond to Tot₁ = (C₁(X,L) ⊗_H C₀(C_ℓ)) ⊕ (C₀(X,L) ⊗_H C₁(C_ℓ)).
Since H = ℤ_ℓ acts freely, the balanced product dimensions are:
dim(C₁(X,L) ⊗_H C₀(C_ℓ)) = |X₁|dim(C₀(X,L) ⊗_H C₁(C_ℓ)) = s|X₀| = 2|X₁|(bys-regularity)
So N = |X₁| + 2|X₁| = 3|X₁|.
The hypothesis hdim_components encapsulates the balanced product dimension formula
dim(Tot₁) = |X₁| + s|X₀|, which follows from the coinvariant dimension formula
dim(V ⊗_H W) = dim(V)·dim(W)/|H| for free actions. The proof then reduces to
regularity: |X₁| + s|X₀| = |X₁| + 2|X₁| = 3|X₁|.
The number of physical qubits is 3|X₁|. Given the balanced product dimensional
decomposition and s-regularity, the proof is arithmetic.
Step 2: Logical Qubit Bound K ≥ (2k_L/s - 1)|X₁|/ℓ #
By Thm_12: K = dim H₁ʰ = dim H₁(C(X/ℤ_ℓ, L)).
By Thm_7 (Sipser–Spielman): dim H₁(C(X/ℤ_ℓ, L)) ≥ (2k_L/s - 1)|X₁/ℤ_ℓ|.
Since |X₁/ℤ_ℓ| = |X₁|/ℓ (free action), K ≥ (2k_L/s - 1)|X₁|/ℓ.
Logical qubit count. The horizontal homology dimension equals the quotient Tanner code homology dimension. This is Thm_12.
Logical qubit lower bound. By Thm_12 (encoding rate) and Thm_7 (Sipser–Spielman),
K = dim H₁(C(X/ℤ_ℓ, L)) ≥ (2k_L/s - 1) · |X₁|/ℓ.
Here kL is the dimension of the local code kernel, numQuotientEdges is |(X/H)₁|
(the number of edges of the quotient graph), and we assume |(X/H)₁| = |X₁|/ℓ
(which holds since H acts freely on edges).
Step 3: D_Z bound #
By Thm_13 (homologicalDistanceBound_horizontal), any nontrivial homology class [x]
with nontrivial horizontal projection satisfies
|x| ≥ |X₁| · min(α_ho/2, α_ho·β_ho/4).
Z-distance bound. Every nontrivial homology class with nontrivial horizontal
projection has weight at least |X₁| · min(α_ho/2, α_ho·β_ho/4). This directly
restates Thm_13, Case 1.
Step 4: D_X bound #
The X-distance combines two cases from Thm_14:
- Case 1 (vertical):
|x| ≥ |X₀|s · min(α_co/2, α_co·β_co/4)Substituting|X₀|s = 2|X₁|:|x| ≥ min(α_co·|X₁|, α_co·β_co·|X₁|/2). - Case 2 (horizontal):
|x| ≥ ℓ · min(α_co/(4s), α_co·β_co/(4s)).
X-distance bound, Case 1 (vertical projection nonzero).
Any nontrivial cohomology class with nontrivial vertical projection satisfies
|x| ≥ |X₀|·s · min(α_co/2, α_co·β_co/4).
By s-regularity |X₀|·s = 2|X₁|, this gives
|x| ≥ min(α_co·|X₁|, α_co·β_co·|X₁|/2).
X-distance bound, Case 2 (purely horizontal).
Any nontrivial cohomology class with zero vertical projection satisfies
|x| ≥ ℓ · min(α_co/(4s), α_co·β_co/(4s)).
Combined D_X bound #
The overall X-distance is the minimum over both cases. For any nontrivial
cohomology class with nontrivial logical projection, we have
|x| ≥ min(|X₀|s · min(α_co/2, α_co·β_co/4), ℓ · min(α_co/(4s), α_co·β_co/(4s))).
Combined X-distance bound. For any nontrivial homology class x with
nontrivial horizontal projection (logical part), the weight is bounded below
by the minimum of the vertical and horizontal cohomological bounds.
Using |X₀|s = 2|X₁|, this gives:
D_X ≥ min(α_co·|X₁|, α_co·β_co·|X₁|/2, ℓ·α_co/(4s), ℓ·α_co·β_co/(4s)).
Satisfiability Witnesses #
Combined distance bound for nontrivial homology classes. Given a nontrivial
homology class in the balanced product code, the combined D_X bound holds. This follows
directly from subsystemCodeParameters_DX_combined.
This replaces the satisfiability axiom: the actual distance bound for any given
nontrivial x follows from the theorems above. The existence of nontrivial homology
classes for specific graph families (Cayley expanders) is a constructive
combinatorial result from [PK22, Section 5].